FailBench

FailBench is a language and reasoning annotation robotics dataset maintained by not stated. Primary hardware platform: 14 public sources, 12 real-world and 2 simulated. Access is not stated under not stated.

Dataset specifications
Year2026
Embodiments14 public sources, 12 real-world and 2 simulated
Task categoriesmanipulation
Licensenot stated
Accessopen
Maintainernot stated